Generative AI increasingly enters scholarly writing as a cognitive and linguistic prosthetic: a tool that extends thought through language while hiding its mediation in fluency. This article introduces the AI Usage Facts label (https://ailabel.netlify.app), a customizable disclosure tool modelled on the nutrition facts label, and argues that disclosure can become a practice of self-knowledge rather than a ritual of compliance. The label asks authors to reconstruct where AI entered the work, how heavily each stage relied on it, and what forms of verification followed. In doing so, it creates deliberate friction: a structured pause in which the writers entanglement with the model becomes perceptible again. The article shows how the labels design choices, including staged workflow reconstruction, Human/Cyborg/AI classifications, a rough percentage index, quality attestations, and live preview, translate a phenomenological account of prosthetic authorship into an interface. It also argues that granular disclosure alters the affective economy of AI use. A detailed account of assistance, reliance, and verification offers readers something more useful than the bare confession I used AI: a basis for calibrated judgment. The article closes by addressing the limits of self-report and the possibility of gaming, arguing that the label cannot guarantee honesty but can cultivate a habit of attention. What begins as a disclosure form may become a small piece of epistemic infrastructure for scholarly writing in the age of cognitive prosthetics.
Faraz Parast (2026) studied this question.